influence of different needle loss symptoms on the
spectral signal is generally lower and is masked by the
influence of varying stand densities. The influence of tree
age and illumination is significant in each band. But by
using synergetic effects of band 4 and 5 this influence
can be reduced.
The separation of different neddle loss categories is not
possible in stands with crown closure less than 60%.
Crown closure below 60% is mostly caused by forest
damage and exeptionally a result of regular forest
management. Therefore, the damage class definition
used in the LAOE is very suitable for describing the
whole range of damage symptoms (opening up and
needle loss).
Classification accuracy
The verification of the results was carried out with
differing methods among the projects. Terrestrical ground
data and aerial photo interpretation was used for the
assessment of accuracy. In the Harz project for example
the accuracy of the classification of the 4 standwise
needle loss classes is about 70%. The combination of the
classes C0, C1 and C2, C3 gave an accuracy of 84%. In
the other test sites the average accuracy of the needle
loss classification was above 89%. The classification of
the three openening up classes can be performed with an
accuracy of over 90%.
of d tru
It has been shown that the methodology of the definition
of representative training areas for the damage classes is
the most crucial point which influencing the classification
result. It is necessary to collect the ground truth data in
such a manner that the stands or training areas can be
directly categorised into the harmonised damage classes.
Harmonisation must already be started by collecting the
ground truth data. The verification areas have to be
defined with the same procedure.
The interpretation of aerial photos is best suite for a
precise definition of ground truth.
9. CONCLUSIONS AND RECOMMENDATIONS
Recent and realistic forest damage maps with
harmonised forest damage classes did not exist at
regional scales until now. The presented project has
demonstrated the value of satellite based remote sensing
data for forest monitoring, in particular for assessing
forest damage areas, where rapid changes in stand
conditions occur.
The satellite based remote sensed data is usable for
forest health status detection. It has been demonstrated
in all the individual investigations of the LAOE that the
reflected radiation recorded by the sensor of Landsat-TM
is affected by changes of forest health status. Satellite
remote sensing has been proven as an appropriate tool
for monitoring forest condition by measuring the spectral
740
characteristics associated with changes in vegetation
health and destruction of forest canopy.
So the classification results can be used as a first phase
of small scale inventories. Additional investigations by the
means of aerial photos or field measurements can be
performed in order to obtain more detailed information on
the level of forest damage. The satellite classification can
provide information for the selection of areas, in which
those detailed inventories have to be carried out.
The digital data sets and maps elaborated in the course
of the LAOE can serve as basis for future change
detection studies and for the installation of a long-term
monitoring system. The results of the project are
comparable among the countries and provide planners at
national and regional level with relevant data for their
decisions. Especially valuable is the ability to control the
effects of already initiated counter measures against the
enormous pollution problems.
With the remote sensing methodology developed in this
project it is generally possible to complement terrestrial
and aerial photography damage assessments. In the
Czech Republic the method is also used operationally for
providing foresters with information they need. on local
level.
The advantages of using satellite data include:
e large area coverage that allows mapping of the
spatial extent of forest damages,
e extended spectral coverage improves measurement
capabilities for assessing the health status of
vegetation,
e archived data sets allow change-over-time studies of
forest stand conditions.
The remote sensing methodology developed in the
project should be brought to the consideration of all
European bodies dealing with the assessment of forest
condition, as a complement to current methodologies.
The methodology should be brought to the attention of
the bodies monitoring the implementation of and in
compliance with the Convention on Biological Diversity as
well as the UNCED Forest Principles, as a complement
to current methodologies.
The results of the project should be brought to the
attention of the governments of Central and Eastern
Europe, the progress of Ministerial Conferences on the
Protection of Forests in Europe, the task force for the
implementation of the Pan-European Biological and
Landscape Diversity Strategy, the CSD International
Panel on Forests and the scientific community dealing
with global change.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996.
Co TT © >
oo —